Premium
Bayesian predictive power for interim adaptation in seamless phase II/III trials where the endpoint is survival up to some specified timepoint
Author(s) -
Schmidli Heinz,
Bretz Frank,
RacinePoon Amy
Publication year - 2007
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.2957
Subject(s) - frequentist inference , interim , interim analysis , bayesian probability , prior probability , sample size determination , computer science , early stopping , bayesian inference , econometrics , statistics , mathematics , randomized controlled trial , machine learning , medicine , artificial intelligence , archaeology , artificial neural network , history , surgery
Integration of a phase II and a phase III clinical trial into a single confirmatory study aims to shorten the development time without compromising the chance of success for a development program. These seamless phase II/III trials involve complex adaptations at the interim analysis, such as treatment selection, sample size reassessment, and stopping for futility. Bayesian methods can support these interim adaptations, and make this decision process more transparent. Use of a frequentist combination test for the final evaluation ensures that the type I error is controlled regardless of the adaptation rule employed at the interim analysis. In this paper, an adaptive seamless phase II/III trial design is proposed for studies where the endpoint is survival up to some specified timepoint and where Bayesian predictive power (PP) guides interim adaptations. For the evaluation of PP at the interim analysis, the event time is modelled as a piecewise exponential distribution, with informative priors for the hazard rates. As an illustrative example, regimen selection at interim in a four‐arm trial with an active control isconsidered, where both non‐inferiority and superiority to the control arm are tested. Frequentist properties of the adaptation criterion based on Bayesian PP are assessed by simulations. Copyright © 2007 John Wiley & Sons, Ltd.